Robinson-Papp, Jessica MD, MS; Elliott, Kathryn MD, JD, MSc; Simpson, David M. MD; Morgello, Susan MD; for the Manhattan HIV Brain Bank
The use of opioids for the treatment of chronic pain in noncancer populations is a controversial subject. This lack of consensus arises from the lack of adequate data supporting the safety and efficacy of opioids in noncancer pain and also from legal and ethical concerns surrounding the harm that may be caused by abuse and diversion of prescription opioids into the community. Thus, the clinician faced with a chronic pain patient, who still seems to be suffering despite nonopioid analgesia, may be conflicted about whether to prescribe opioids. Undertreatment of pain has been documented in the medical literature and publicized in the lay press.1 Efforts to improve the treatment of pain led to the numeric pain score being adopted as the “fifth vital sign.”2 Adequate treatment of pain has been referred to as a “moral imperative.”1,2 Patients claiming that their pain was not adequately addressed have won malpractice suits.3 However, it is increasingly clear that an overly liberal approach to the prescription of opioids is not desirable either. There has been a dramatic rise in prescription opioid abuse and death due to overdose.4 Clinicians have been convicted on criminal charges for overprescription of opioids.5
In light of these mixed messages, there is great interest in determining the most appropriate and safe opioid prescribing practices. To this end, it is desirable to define what constitutes a poor outcome of treatment with opioids. Possible poor outcomes include lack of efficacy, side effects, development of addiction, and diversion of opiates. The former 2 outcomes are similar to those that must be considered for any medication, whereas the latter 2 are more specific to substances with abuse potential.
A body of literature beginning in the late 1990s began addressing addiction and diversion behaviors in patients prescribed opioids, using the terms problematic opioid use, aberrant opioid use, and opioid misuse.6–8 Although there is not a single gold standard definition of problematic use, a simple and broad conceptual definition is use of the prescription opioid in a manner that deviates from the provider's instructions. This includes abuse, misuse, and diversion. Because these are not often directly observable behaviors, more specific clinical features have been described, including unexpected toxicology results, frequent requests for dose increases, failure to adhere to concurrently recommended treatments, frequently reported loss of prescriptions or medications, tampering with prescriptions, prescriptions obtained from multiple providers (including emergency rooms), and missed and/or extra follow-up visits.9
Researchers have sought to determine risk factors for problematic use and to develop tools to predict it. However, there are significant methodological hurdles, as described by Turk et al6 in a systematic review of the literature published in 2008. One of the main difficulties is choosing a valid outcome measure. The researcher is unlikely to be able to objectively document problematic behaviors such as tampering with prescriptions or opioid diversion. The available alternatives include patient self-report, provider report, and urine toxicology screening. Each has significant limitations. Patients have incentive to conceal substance misuse and have been shown to do so.10 Providers have no way of truly knowing what their patients do with their medications. Urine toxicology has the advantage of objectivity but gives a circumscribed picture.
Despite these difficulties, certain patient characteristics have been associated with problematic use in one or more studies and are generally considered to be possible risk factors. The most consistent and perhaps most obvious is a history of drug or alcohol abuse.11–13 Other potential risk factors that have been studied include younger age,13,14 male sex,13,15,16 psychiatric comorbidity,11,12,14,15,17,18 lower education level,18 family history of substance use,11,15 and history of sexual abuse.18
These potential risk factors pose a dilemma for providers treating HIV-infected individuals because many of them are so common in certain HIV populations as to make meaningful risk stratification impossible. Yet pain is also extremely common among HIV-infected individuals. In a recent survey of HIV primary care practices, providers reported on average that 28% of their patients had chronic pain.19 Recent patient-based studies have reported prevalence of pain at more than 50%.20–22 Prescription of opioids to HIV-infected individuals is also higher than in the general population.23
A few publications have begun to examine problematic use specifically in HIV-infected individuals.24–26 Hansen et al25 studied a group of marginally housed or homeless HIV-infected adults in San Francisco and found a high prevalence of problematic use, as assessed by an affirmative answer to at least one of 20 items queried using an audio computer-assisted self-interview. Tsao et al26 used data from the HIV cost and services utilization study (HCSUS) and found greater opioid misuse among patients with histories of drug problems. Both variables were defined based on responses to questionnaire items developed for HCSUS. However, a study of a broad range of risk factors for problematic prescription opioid use in HIV-infected individuals has not been performed. In this study, we sought to determine whether the risk factors described among chronic pain patients would also be associated with problematic use in HIV-infected individuals, using both urine toxicology and psychiatric interview to define outcomes. In addition, we sought to examine the role of some factors specifically pertinent to HIV, namely adherence to antiretrovirals (ARVs), CD4+ count, plasma HIV viral load (VL), and burden of comorbid illnesses.
Participants and Study Procedures
Data were obtained from the Manhattan HIV Brain Bank (MHBB), an ongoing prospective cohort study established in 1998 to provide a resource of nervous system tissues from HIV-infected donors who undergo comprehensive, longitudinal, antemortem evaluation. The MHBB is a member of a 4-site consortium, the National NeuroAIDS Tissue Consortium. The primary enrollment criterion for the MHBB is willingness to participate in organ donation at the time of death. Other criteria were designed to target patients with advanced disease (eg, hemoglobin less than 10 g/dL, CD4 count less than 50, and opportunistic infections) or other illness refractory to therapy (congestive heart failure, end-stage liver disease). After enrollment, participants are evaluated at 3- to 6-month intervals until the time of death, with sicker participants evaluated more frequently than healthier participants. Assessments include collection of demographic data, standardized medical interview, general physical and comprehensive neurological examination, and assessment of Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) diagnoses of psychiatric and substance use disorders using the Psychiatric Research Interview for Substance and Mental Disorders (PRISM). Laboratory evaluations include CD4+ count, plasma VL, and urine toxicology for substances of abuse (amphetamines, benzodiazepines, barbiturates, cannabinoids, cocaine, methadone, opiates, phencyclidine, and propoxyphene). Data were obtained under a protocol approved by the institutional review board of the Mount Sinai School of Medicine. All research procedures were in accordance with the Helsinki Declaration of 1975, as revised in 2000.
All MHBB participants who were prescribed an opioid, other than methadone, were considered in the analysis. Methadone was excluded because in the vast majority of our participants, it is prescribed as methadone maintenance treatment for addiction rather than for pain management. Because data from multiple visits were available for most patients, the first visit at which the opioid appeared on the prescription medication list was used. Problematic opiate use was defined as a positive urine toxicology for substances not prescribed by a physician (amphetamines, benzodiazepines, barbiturates, cannabinoids, cocaine, methadone, opiates phencyclidine, and propoxyphene), or a current substance abuse or dependence diagnosis for any substance (alcohol, cannabis, cocaine, hallucinogens, opiates, sedatives, stimulants or other substances), as determined by the PRISM. The following potential risk factors for problematic use were available and were considered in the analysis: age, gender, ethnicity, years of education, literacy (as determined by the Wide Range Achievement Test 3 reading subscale; WRAT), past substance abuse or dependence, the presence of any nonsubstance-related psychiatric disorder either current or past [with specific attention to major depressive disorder (MDD), generalized anxiety disorder (GAD), and posttraumatic stress disorder (PTSD)], CD4+ count, plasma VL, whether ARVs were prescribed, adherence to ARVs, and burden of comorbid illnesses as measured by the Charlson comorbidity index. All psychiatric and substance use diagnoses were determined using the PRISM.
We sought to make the best use of the available data and so both cross-sectional and longitudinal analyses were performed, as illustrated in Figure 1. In cross-sectional analyses, univariate associations (t test, Wilcoxon rank sum, or χ2, as appropriate) were sought between problematic use and age, gender, ethnicity, years of education, WRAT, CD4+ count, plasma VL, whether ARVs were prescribed, adherence to ARVs, Charlson comorbidity index, PRISM diagnosis of past substance abuse or dependence, any nonsubstance-related psychiatric disorder (current or past considered separately), MDD, GAD, and PTSD. Factors associated with problematic use at a significance level of P < 0.2 were considered for inclusion into a multivariate logistic regression model with problematic use as the dependent variable. A backward selection technique was used to find the best model. The subset of participants who did not have problematic use at baseline and had at least one follow-up visit, at which they were still prescribed opioids, were included in longitudinal analyses. Cox proportional hazards models were used to examine risk factors for future problematic use. We also identified an equal number of contemporaneous participants who were not exposed to prescription opioids and used similar models to determine if they differed from the opioid-prescribed group in rates of future substance misuse. The members of this “unexposed” cohort were chosen to match members of the “exposed” cohort for date of entry into the study and visit number. This was done to allow each group to have similar potential follow-up time. In addition, the unexposed group was enriched for participants with a history of substance use in an attempt to have comparable proportions of former substance users in both groups, although individuals were not matched on this basis.
The characteristics of the cross-sectional sample are described in Table 1. Overall, the participants were diverse with regard to gender and ethnicity. Most had long-standing HIV and were prescribed ARVs. Substance use history and psychiatric comorbidity were common. Oxycodone and codeine, including combination preparations such as Percocet or Tylenol 3, were the most commonly prescribed opioids.
Cross-Sectional Univariate Analyses
Problematic opioid use was very common in cross section, 64% of participants met the definition, that is, they had a positive urine toxicology for a nonprescribed substance or a current PRISM abuse or dependence diagnosis. Among participants with problematic use, the most common substance of abuse was cocaine (42%), followed by cannabis (26%), nonprescribed opiates (26%), alcohol (11%), and sedatives (11%). Stimulants and hallucinogens were each used by one patient. Thirty-four participants (20%) were using more than one substance. Of note, there was significance discordance between the results of the urine toxicology and the PRISM. Among the participants with an illicit substance detected by urine toxicology, only 26% reported substance use on the PRISM. This percentage varied by the individual substance (eg, 37% for cocaine, 19% for cannabis, and 10% for nonprescribed opiates) but small numbers and participant overlap between the groups precluded formal comparison.
The results of univariate analyses (t test, Wilcoxon rank sum, or χ2, as appropriate) performed to explore associations between problematic opioid use and potential risk factors are shown in Table 2. Current nonsubstance-related psychiatric disorder was strongly associated with problematic opioid use (P = 0.001). Among those with a current psychiatric disorder, 78% had problematic use compared with 53% among those without a current psychiatric disorder. Three specific diagnoses were investigated: MDD, PTSD, and GAD. Problematic use was most closely associated with PTSD (P = 0.001), followed by MDD (P = 0.02) and GAD (P = 0.09). A past substance-use diagnosis was also associated with problematic use (P = 0.04). Among past substance users, 67% had problematic use, compared to 46% of those with no past substance diagnosis. Among the participants who were prescribed ARVs, poorer adherence (<95%) was associated with problematic use. Among those with poorer adherence, 80% had problematic use, compared with 58% of those who consistently took their ARVs.
Cross-Sectional Multivariate Analyses
Multivariate logistic regression using backward selection was performed with the following variables initially entered into the model: age, past nonsubstance-related psychiatric disorder, current nonsubstance-related psychiatric disorder, past substance-use diagnosis, and log (VL). Two variables remained in the model (P = 0.001): current nonsubstance-related psychiatric disorder [odds ratio (OR) = 3.1; 95% confidence interval (CI): 1.5 to 6.2] and past substance-use diagnosis (OR: 2.5; 95% CI: 0.98 to 6.2).
Adherence was not included in the above analysis because it was not relevant for all patients because some were not prescribed ARVs. However, the same regression as described above with the additional variable of adherence was performed on the subset of participants prescribed ARVs. Two variables remained in the model (P = 0.009): current nonsubstance–related psychiatric disorder (OR: 2.5; 95% CI: 1.0 to 6.0) and poorer adherence (OR: 3.55; 95% CI: 1.3 to 9.5).
Sixty-two participants did not meet criteria for problematic opioid use in the cross-sectional analysis. Of these, 35 had at least one follow-up visit at which they were still prescribed opioids and were therefore suitable for inclusion in the longitudinal analysis. Among these 35 participants, 10 (29%) had a current nonsubstance-related psychiatric disorder and 28 (80%) had a past substance-use diagnosis. The median time of follow-up was 16 months (interquartile range: 10–41). Thirteen (37%) patients displayed problematic use during follow-up. Due to small numbers, only the 2 strongest variables from cross-sectional analysis were examined in Cox proportional hazards models, current nonsubstance-related psychiatric diagnosis and past substance-use diagnosis. Neither risk factor predicted problematic use (P > 0.8 for both variables) (Fig. 2).
The characteristics of the prescription opioid exposed and unexposed groups are described in Table 3. The groups were similar except that the exposed group had a higher burden of comorbid illnesses and a trend for lower literacy. In the unexposed group, 10 (29%) participants had evidence of substance misuse during follow-up. Cox proportional hazards models adjusting for Charlson score and WRAT showed no difference in future substance misuse between the prescription opioid exposed and unexposed (P = 0.91).
In this study, we sought to explore risk factors for problematic prescription opioid use in a cohort of HIV-infected adults. Our cross-sectional analysis of 173 participants demonstrates an association between problematic use and current psychiatric disorder, history of substance abuse or dependence, and poor adherence to ARVs. These findings are in keeping with previous literature, most of which was also cross-sectional in design. However, an inherent limitation in the interpretation of cross-sectional data is the inability to make inferences about causality. For example, it could be argued that the association of past substance-use disorders and problematic use simply demonstrates that those who have abused substances in the past are more likely to be doing so currently and the fact that an opioid has been prescribed is irrelevant and just an artifact of the study design. Another potential limitation of the cross-sectional analysis is its clinical relevance. According to some experts, urine toxicology should be performed before an opioid is prescribed and patients found to be actively using illicit substances should usually not be prescribed opioids. Accordingly, longitudinal analyses that were more relevant to the clinical practice recommended by these guidelines were performed. When we followed participants who did not have problematic use at baseline, we found that although a sizable percentage (37%) went on to develop problematic use, past substance use and psychiatric disorder at baseline were not predictive. We also found that participants prescribed opioids were not more likely to develop evidence of a substance use disorder during treatment than participants who were not prescribed an opioid. These findings suggest that although problematic use is associated with active psychiatric illness and past substance use at a given point in time, these factors cannot be used to identify patients who will have problematic use in the future. Furthermore, treating these “high-risk” patients with opioids does not seem to increase the rate at which they abuse substances in the future.
How might the discrepancy between cross-sectional and longitudinal data be explained? One explanation is the loss of statistical power from the larger cross-sectional sample to the small longitudinal group. However, it is notable that the groups with and without “risk factors” showed very similar cumulative incidence of problematic use (Fig. 2), as demonstrated by the very high P values (>0.8). This makes the existence of a true difference less likely even when given the very small sample size.
An alternative explanation is that although patients actively abusing substances are likely to have active psychiatric disorders, past substance-use disorders and poorer adherence to ARVs, the presence of these factors does not predict future substance misuse in this complicated population. Current guidelines for the general chronic pain population emphasize the importance of risk assessment when considering opioids for treatment of chronic pain.27 Our findings suggest that traditional factors used to assess risk may not be valid in HIV-infected individuals. Thus, using these factors when weighing a decision about who to prescribe opioids to might unjustly deprive some patients of a helpful therapy and give the provider a false sense of security when prescribing opioids to other patients. Our findings also have relevance to the use of opioid contracts. We found a significant discordance between patient's report of substance use behaviors and the results of urine toxicology, which demonstrates that participants are often not truthful when discussing substance use. This suggests that a patient's agreement to an opioid contract may not be meaningful, especially, when they must agree to receive opioids.
There are important limitations to this study. Most of the participants have long-standing HIV and/or significant comorbid illnesses. This may limit generalizability. The data were obtained from an ongoing cohort study not specifically designed for the purpose of this analysis, thus information about problematic behaviors other than illicit substance use, such as diversion, were not available. Similarly, we did not have detailed urine toxicology data, capable of distinguishing different opioids, or detailed data about the specific indication for the prescribed opioid. Another limitation is the small number of opioid-exposed patients with longitudinal data. However, despite these limitations, this study provides important information because it is the first to examine a breadth of risk factors for problematic prescription opioid use in HIV-infected individuals.
In summary, our findings demonstrate that problematic prescription opioid use is very common in HIV-infected individuals but cannot be readily predicted based on patient characteristics, and furthermore, that self-report of problematic use is not reliable. We do not wish to suggest that opioids are never appropriate for chronic pain in HIV-infected individuals or that the report of chronic pain should not be taken seriously. Quite the contrary, we have reported previously that in the example of neuropathic pain, self-reported symptoms were highly predictive of an objective neurologic abnormality, namely distal symmetric polyneuropathy.28 However, when contemplating opioid treatment for chronic pain in HIV-infected individuals, the provider should be aware that there is a relatively high likelihood that problematic use will develop. Based on the results of this study, we recommend that providers prescribing opioids to HIV-infected individuals avoid opioid contracts and attempts at risk factor assessment and instead adopt “universal precautions,” including baseline and follow-up urine toxicologies, accompanied by a frank conversation emphasizing that unexpected results will lead to discontinuation of opioid treatment.
The authors acknowledge and thank the participants and staff of the Manhattan HIV Brain Bank.
1. Institute of Medicine Committee on Advancing Pain Research, Care, and Education. Relieving Pain in America: A Blueprint for Transforming Prevention, Care, Education, and Research. Washington, DC: The National Academies Press; 2011.
2. Stephenson J. Veterans' pain a vital sign. JAMA. 1999;281:978.
4. Warner M, Chen LH, Makuc DM. Increase in fatal poisonings involving opioid analgesics in the United States, 1999-2006. NCHS Data Brief. 2009;22:1–8.
6. Turk DC, Swanson KS, Gatchel RJ. Predicting opioid misuse by chronic pain patients: a systematic review and literature synthesis. Clin J Pain. 2008;24:497–508.
7. Jamison RN, Butler SF, Budman SH, et al.. Gender differences in risk factors for aberrant prescription opioid use. J Pain. 2010;11:312–320.
8. Robinson RC, Gatchel RJ, Polatin P, et al.. Screening for problematic prescription opioid use. Clin J Pain. 2001;17:220–228.
9. Ballantyne JC, Mao J. Opioid therapy for chronic pain. N Engl J Med. 2003;349:1943–1953.
10. McNagny SE, Parker RM. High prevalence of recent cocaine use and the unreliability of patient self-report in an inner-city walk-in clinic. JAMA. 1992;267:1106–1108.
11. Michna E, Ross EL, Hynes WL, et al.. Predicting aberrant drug behavior in patients treated for chronic pain: importance of abuse history. J Pain Symptom Manage. 2004;28:250–258.
12. Schieffer BM, Pham Q, Labus J, et al.. Pain medication beliefs and medication misuse in chronic pain. J Pain. 2005;6:620–629.
13. Ives TJ, Chelminski PR, Hammett-Stabler CA, et al.. Predictors of opioid misuse in patients with chronic pain: a prospective cohort study. BMC Health Serv Res. 2006;6:46.
14. Edlund MJ, Martin BC, Fan MY, et al.. Risks for opioid abuse and dependence among recipients of chronic opioid therapy: results from the TROUP study. Drug Alcohol Depend. 2010;112:90–98.
15. Liebschutz JM, Saitz R, Weiss RD, et al.. Clinical factors associated with prescription drug use disorder in urban primary care patients with chronic pain. J Pain. 2010;11:1047–1055.
16. Edlund MJ, Steffick D, Hudson T, et al.. Risk factors for clinically recognized opioid abuse and dependence among veterans using opioids for chronic non-cancer pain. Pain. 2007;129:355–362.
17. Becker WC, Sullivan LE, Tetrault JM, et al.. Non-medical use, abuse and dependence on prescription opioids among U.S. adults: psychiatric, medical and substance use correlates. Drug Alcohol Depend. 2008;94:38–47.
18. Wasan AD, Butler SF, Budman SH, et al.. Psychiatric history and psychologic adjustment as risk factors for aberrant drug-related behavior among patients with chronic pain. Clin J Pain. 2007;23:307–315.
19. Lum PJ, Little S, Botsko M, et al.. Opioid-prescribing practices and provider confidence recognizing opioid analgesic abuse in HIV primary care settings. J Acquir Immune Defic Syndr. 2011;1(suppl 56):S91–S97.
20. Lee KA, Gay C, Portillo CJ, et al.. Symptom experience in HIV-infected adults: a function of demographic and clinical characteristics. J Pain Symptom Manage. 2009;38:882–893.
21. Merlin JS, Cen L, Praestgaard A, et al.. Pain and physical and psychological symptoms in ambulatory HIV patients in the current treatment era. J Pain Symptom Manage. 2012;43:638–645.
22. Dobalian A, Tsao JC, Duncan RP. Pain and the use of outpatient services among persons with HIV: results from a nationally representative survey. Med Care. 2004;42:129–138.
23. Silverberg MJ, Ray GT, Saunders K, et al.. Prescription long-term opioid use in HIV-infected patients. Clin J Pain. 2012;28:39–46.
24. Passik SD, Kirsh KL, Donaghy KB, et al.. Pain and aberrant drug-related behaviors in medically ill patients with and without histories of substance abuse. Clin J Pain. 2006;22:173–181.
25. Hansen L, Penko J, Guzman D, et al.. Aberrant behaviors with prescription opioids and problem drug use history in a community-based cohort of HIV-infected individuals. J Pain Symptom Manage. 2011;42:893–902.
26. Tsao JC, Stein JA, Dobalian A. Pain, problem drug use history, and aberrant analgesic use behaviors in persons living with HIV. Pain. 2007;133:128–137.
27. Chou R, Fanciullo GJ, Fine PG, et al.; American Pain Society-American Academy of Pain Medicine Opioids Guidelines Panel. Clinical guidelines for the use of chronic opioid therapy in chronic noncancer pain. J Pain. 2009;10:113–130.
28. Robinson-Papp J, Morgello S, Vaida F, et al.. Association of self-reported painful symptoms with clinical and neurophysiologic signs in HIV-associated sensory neuropathy. Pain. 2010;151:732–736.
© 2012 Lippincott Williams & Wilkins, Inc.